In the days of Folksonomy, it is noce to see researchers report interesting progress that could lead to automatic tagging and image feature based searches rather than human assigned tag only. A recent article describes success with Gesture recognition work by Thomas Huang of University of Illinois and Nicu Sebe of University of Amsterdam (using among others Mona Lisa). Tom has been my hero since my student days and it is amazing to see that he continues his work with same dedication and passion even today. A true insipration.
Most commercial systems talk about manual tagging and use only text for searching image. Academia has been working on this research area for quite some time and with some creative efforts, I feel, that now it has technology that could in fact be very useful and applicable to search on Web. It is amzing that 10 years ago, there were many projects and systems on the Web to search images using ‘content’ of images. At least three companies ( Virage, IBM, and Excalibur ) were deveoping Web based image search. Now 10 years later, we have very little on this in commercial world.
I think time has come for some entrepreneurs to look into this.